I am looking for a model, characterized continuously by a single parameter, to describe the arrival times of buses with unit expected interarrival time. At one extreme of the parameter (say $\theta=1$), the process is deterministic, with all interarrival times equal (to $1$). When $\theta=0$, the process is pure Poisson (with unit expected interarrival or waiting time). As $\theta$ approaches $-1$, the arrivals tend to cluster, with long intervals between the clusters. At the (unattainable) extreme of $\theta=-1$, all the buses arrive in a simultaneous convoy, after an infinite wait, and you have to wait forever for the next convoy.

The choice of $\{1,0,-1\}$ for the extreme and central parameter values isn't important: they could be $\{-\infty, 0, \infty\}$, $\{0,\frac12,1\}$, ..., whatever. At this stage, simplicity and naturalness, subject to the above conditions, is more important than realism.


In a pure Poisson process, the interarrival times (times between each bus) has an exponential distribution. So you want some family of distributions which includes the exponential as a special case. The gamma distribution (see: https://en.wikipedia.org/wiki/Gamma_distribution). When then shape parameter is 1, you have the exponential. When the shape parameter ($k$ or $\alpha$ in wiki) approaches zero, you are approaching constant interarrival times, and growing above 1 you get interarrival times more variable than for the poisson process.

| cite | improve this answer | |
  • $\begingroup$ Yes, this could be the model if we set $k\theta=1$. But I think that you have it the wrong way round: We approach a constant (unit) interarrival time as $k\rightarrow\infty$, and long intervals between clusters as $k\rightarrow0$. $\endgroup$ – John Bentin Feb 12 '15 at 16:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.